transformers
cc2237c5
- Simplify parallel plan resolution and clarify TP and EP meshes
Go
Login via GitHub
Home
Pricing
FAQ
Install
Login
via GitHub
Commit
View On
GitHub
Commit
4 days ago
Simplify parallel plan resolution and clarify TP and EP meshes
References
#48204 - [Umbrella] FSDP2 + expert-parallel token dispatch, with sharded-loading prefetch
#48792 - Add Expert parallel
Author
3outeille
Parents
c8a05bc6
Loading